Metal Forming - Hot Rolling

Understanding the Properties of Oxide Scales on Hot Rolled Steel Strip

P. Henk Bolt

Corus Research, Development & Technology, ljmuiden I The Netherlands

Surface quality and downstream processability of hot rolled strip depend on tertiary scale thickness and composition at all process stages,
from secondary descaling to the cold coil. TGA oxidation experiments provided the basis for quantitative modelling of scale growth during
finishing hot rolling of low carbon steels (with or without Si or Pl. Finishing temperature (FT), gauge, and (particularly at the edges) coiling
temperature (CT) are the main influence factors with respect to scale thickness. Based on HTXRD oxidation trials and microscopic analy-
ses, trendlines are presented that roughly predict the final scale composition as a function of CT, for the coil centre and the edges. It ap-
pears that a CT of 400-600°C, in combination with a coil centre position, is favourable for decomposed wustite, which is the preferred com-
ponent for a good scale ad- and cohesion. Retention of wustite, the most preferred phase in the roll bite but very brittle at room tempera-
ture, is promoted by relatively high CTs (particularly with a coil quench) and by very low CTs. Hematite is formed at the edges in case of a
high CT. This extremely hard phase is unwanted at any stage.

Keywords: hot rolled steel strip, oxide scale, TGA, HTXRD.

Introduction M = J2. k p ' t (3)

During finishing rolling of hot rolled steel strip, tertiary
oxide scale is developing. Its properties, depending on
process conditions and steel chemistry, have a large impact
on events in the roll bite, and hence on surface quality.
Moreover, scale properties determine the processability of a
hot rolled coil in downstream processes like pickling, laser
cutting, forming, stamping, or direct coating. Relevant
properties of the oxide layer are thickness, composition,
structure, adhesion and cohesion.
Scale growth. The thickness of any oxide layer depends
on the time-temperature cycle, rolling reduction scheme, al-
loy composition and atmosphere conditions. When the ox-
ide layer is still rather thin, the scale growth rate will either
be determined by gaseous transport of Oz towards the sur-
face (or another oxidising species like HzO or COz), or by
the chemical reaction itself.
In such a case, the scale growth rate dM/dt is independent
of scale thickness M as indicated in figure 1. This is de-
scribed by the linear rate equation [1, 2]:
which is actually the integrated form of the more funda-
mental equation:
dM kp
= (4)
dt M
Here k p is the parabolic rate constant.
Although equations (3) and (4) are not valid for thin ox-
ide layers, when Fez+ diffusion is not yet rate-limiting, it is
quite popular to assume simple parabolic growth kinetics
from the start, since it is a relatively simple way to describe
scale growth. This is allowable for long oxidation periods,
like in reheating furnaces, where the initial non-linear stage
can be neglected. However, also in papers dealing with rel-
atively short oxidation time periods, like secondary and ter-
tiary scale growth in a hot strip mill, simple parabolic be-
haviour is frequently assumed. This is incorrect. As illus-
trated in figure 1, simple parabolic kinetics imply scale

(1) 20

::i 15
~
Here ki. is the linear rate constant. In integrated form •
(with kdT) = constant, i.e. under isothermal conditions I:
~ 10
within the time period t): ...
ii
~
•
M = «c t (2) •
5

When the oxide layer is rather thick, solid-state diffusion

of Fez+ will be rate-limiting. The Fez+ diffusion rate de- 30 60 90 120
creases with increasing scale thickness (i.e. diffusion dis- time (a.u.)

tance). This situation is commonly described by the well- Figure 1. Four types of scale growth kinetics (under isothermal con-
known simple parabolic rate equation [1, 2]: ditions).

steel research into 75 (2004) NO.6 399

Metal Forming - Hot Rolling
growth rates approaching infinity at t...(), which is physical-
ly impossible. The mixed parabolic rate equation is a better
description of the oxidation kinetics, since it takes both the
linear and the parabolic stage into account [I, 2]:
dM kp kp
- - --- with ex = - (5)
dt - M +ex k:
For M«a, equation (5) reduces to equation (1), and for
M»a, equation (5) reduces to equation (4). Figure 1 con-
firms that the relative difference in predicted scale thickness
between simple and mixed parabolic rate equations (with
the same kp ) becomes increasingly unimportant for large t
and M. At small t and M values, however, the simple para-
bolic rate equation becomes very inaccurate. Figure 1 also
illustrates a fourth type of oxidation kinetics, namely para-
linear kinetics. In some cases, it is found empirically that
the oxide layer is protective only up to a certain thickness
M L. When ML is exceeded (M=M L+L1M), the extra scale 11M
is non-protective. In such a case, transport of gaseous com-
ponents like O 2, H20 or CO 2 is possible through this sub-
layer. As a result, the kinetics of further scale growth are de-
termined by Fe2+ diffusion through the part of the scale that
is still protective, with thickness M L :
dM
- = kc f at M > ML (6)
dt '
Here kLj is the final linear rate constant; kLj < kL.
TG A experiments
Quantitative modelling of secondary and tertiary scale
growth on low-alloy steels implies assessment of numerical
expressions for kp, kL, and possibly kLj as functions of tem-
perature and alloy composition. To this end, a series of ox-
idation experiments by thermal gravimetry anaysis (TGA)
were performed in a Cahn TG 171 thermobalance, at tem-
peratures between 700 and 12OO°C. Si and P had been iden-
tified as the most important alloying elements affecting
scale growth on low-alloy steels. Therefore, LC steel sam-
ples with Si levels up to 0.4 wt.% and P levels up to 0.1
wt.% were oxidised, in dry or humid air (0-35 vol.% H20
vapour). Figure 2 shows a few examples of the large col-
lection of obtained TGA curves [3].
From the TGA experiments it appeared that:
- Scale growth is retarded by both Si and P (figure 2). This
effect is relatively strong at lower oxidation temperatures.
- Paralinear kinetics were particularly observed with P-al-
loyed steels. This is associated with a tendency to form
blisters and cracks within the scale, which makes the
scale partially non-protective (inset of figure 2). Indeed,
industrial P-alloyed hot rolled steels also show a poor
scale coherence, with a relatively high porosity and crack
level.
- Scale growth in humid air is significantly faster than in
dry air (figure 2). Moreover, the adherence of scale
formed in humid air was significantly better than that of
400
0,3)
0,25
l;0,20
~
i 0,15
LC steel with 0.4% Si,
II
at
humid air (15 vol%)
5°,10 -,
...............¢. ,,~ -.;
E i LC steel with l
0,05 l0.1% P, humid ~ Scale blistering
lair (15 vol%)
~_ "' _
; .
1 paralinear kineticf
120 240 360 600
time (s)
Figure 2, TGA curves recorded during oxidation of low carbon
steels at 1000°C in dry and humid air.
scale formed in dry air. Apparently, scale constitution de-
pends on humidity. This implies that laboratory oxidation
experiments performed with dry air have little predictive
value for the humid conditions in a hot strip mill.
- Scale growth is strongly temperature-dependent. This can
be described by Arrhenius expressions,
(7)
(8)
The influence of Si, P and humidity levels can be de-
scribed by expressions for kto. QL, kro. Qp and kLj as a func-
tion of these parameters. Parabolic activation energies Qp
range from 175.6 kl/mole for low-P steel to 197.6 kl/mole
(humid air) or 213.4 kl/mole (dry air) for high-P steel. The
linear activation energy QL is 60.7 kl/mole for the base al-
loy, but increases with increasing P and Si levels.
Tertiary scale thickness
Tertiary scale growth during finishing hot rolling can be
modelled with the obtained expressions, in combination
with the alloy composition, time-temperature cycle and
rolling reduction scheme of the strip as indicated in
figure 3.
It should be pointed out that in this respect, surface tem-
perature Tsurf is the relevant temperature. Tsurf can be sub-
stantially below the average strip temperature, particularly
at and just after locations where the strip is cooled, i.e, dur-
ing descaling, interstand cooling, and in contact with the
cooled work rolls.
The most important parameters with respect to scale
thickness at the exit side of the finishing train are fmishing
temperature (FT) and gauge. Further scale growth occurs on
the run-out table (ROT), depending on the amount of cool-
ing applied, and thus on the coiling temperature (CT). How-
ever, the amount of scale grown on the ROT is modest since
temperatures are appreciably below the temperatures pre-
vailing upstream (Figure 4). The edges are somewhat cold-
er than the strip centre, so up to the coiler, edge scales are
thinner than centre scales. This is unchanged after coiling at
steel research int. 75 (2004) No.6
 Metal Forming - Hot Rolling

E 12
14 -------------------------1 LowCT High CT
• scale
formed
after
:a. coilling
-; 10 .scale
!
.!Ill
U
8
formed
on ROT
5 6
.!! escale
rJ 4 coller
formed
III during
2 rolling
centre edge centre edge
O¥'---.-----.,.-----,-----.---r----.,.----'
o 5 10 15 20 25 30
time (8) Figure 4. Schematic representation of the relative contribution of
process stages (finishing hot rolling, run-out table, and coil cooling)
to tertiary scale growth.
Figure 3. Calculated thickness of scale, developing in a hot strip fin-
ishing mill.

.~'-----'.

(a) . upper ' (b) i upper \

a low CT (roughly below 600°C). However, a substantial
amount of extra scale growth occurs at the edges after coil-
ing at a high CT (roughly above 650°C), due to air ingress
between the windings. As a result, edge scales are thicker
than centre scales at a high CT, figure 4.
The final scale thickness is relevant for throughput and
mass balance of pickling lines. Moreover, thin scales can
generally withstand higher loads than thick scales before
failure (cracking) occurs. This is of importance for the ad-
hesive and cohesive properties of the final scale on unpick-
led hot rolled strip.
Scale in the roll bite
This also applies to the intermediate scale thickness, dur-
ing rolling reductions. Figure 3 clearly shows that the oxide
layers entering the first three finishing stands, FI-F3, are
usually considerably thicker than the final scales. The scale
at these intermediate stages should not be too thick, other-
wise the scale is cracked in the roll bite, rather than plasti-
cally deformed, which results in a poor surface quality,
figure S.
Whether Figure 5a or 5b prevails does not only depend on
the incoming scale thickness and the rolling reduction.
Scale temperature at the entry of a roll bite is also an im-
portant factor, since relatively cold scales are less de-
..:.... ..:....
i work 'I I work "
Figure 5. a) Plastic deformation of scale in the roll bite (desired); b)
scale cracking (undesired).
formable than hot scales. Therefore, strip cooling (during
descaling and interstand cooling) should be carefully bal-
anced. Too much cooling at the wrong moment causes over-
chilling of the scale, whereas insufficient cooling leads to
high surface temperatures, and thus excessive scale growth.
Moreover, surface temperatures above 950°C affect the
scale composition adversely. For the events in the roll bite,
scale composition is even more important than scale thick-
ness.
Scale composition
Three different iron oxide phases can be found in the
scale, namely Fel_xO (wustite), Fe304 (magnetite) and
FeZ03 (hematite). Table 1 lists some properties of these ox-
ides. In particular wustite has some extraordinary features.
Firstly, it has an extremely broad composition range; x val-
ues in Fel_xO range from 0.04 to 0.17. Secondly, wustite is

Table 1. Components in oxide scales, and some of their properties.

Hot I) Tensile Effect on

Vickers Strength 2) Pickling
Name Formula Fe/O ratio scale
Deformability Hardness 2) rate
(MPa) lntearity 2)
Hematite Fe201 0.67 bad ± 1000 10 bad very slow
Magnetite Fe104 0.75 fair ±550 40 fair slow
Wustite Fe1_xO 0.83 - 0.96 good ±300 0.4 bad fast
Decomposed
Fe304 + a-Fe 0.83 - 0.96 N.A. ± 550 >40 good slow
lWustite
I) under hoi rolling conditions 2) at room temperature

steel research into 75 (2004) No.6 401

Metal Forming - Hot Roiling

thermodynamically unstable below 570°C. During cooling, Of course, step 4 was by far the most time-consuming
reaction (9) occurs, yielding "decomposed wustite"; i.e. a step, whereas step 3, the actual oxidation step, was relative-
mixture of finely distributed a-Fe particles in a magnetite ly short. HTXRD diffractograms were recorded during
matrix: these steps (and in some cases also at the end of step 2), ini-
tially at a high frequency (1-3 per minute), then at lower fre-
(9) quencies (1-6 per hour). A large series of HTXRD experi-
ments was carried out to investigate the influence of simu-
In many cases, less than 100% of the wustite is trans- lated FT, CT, position within the coil (centre versus near-
formed following reaction (9); therefore retained wustite is edge region), and coil cooling conditions (water-quench
often observed in oxide scales at room temperature. versus unforced cooling). Figure 6a shows an example of a
At hot rolling temperatures, wustite exhibits a remarkably series of HTXRD patterns recorded during an experiment
high plasticity, but wustite is extremely brittle at room tem- with simulated FT=760°C, CT=620°C, coil centre, un-
perature. So wustite is the desired oxide phase during quenched. For comparison, figures 6b and 6c show the ef-
rolling, but high fractions of retained wustite in the cold coil fects of coil quenching, and of a rather low coiling temper-
are detrimental to the cohesion of the scale. ature of 380°C, respectively.
Hematite is brittle and extremely hard, and therefore un-
desirable at any stage of the process route. In the roll bite,
hematite causes excessive roll wear and a bad strip surface
quality. 40 42 44 46 48 50
Magnetite, with an intermediate Fe/O ratio and an inter- (8)
mediate hardness, shows some plasticity at hot rolling tem- FT=760'C
CT=620'C
peratures. Therefore, a certain fraction of magnetite can be coilcentre
tolerated during rolling reductions. At room temperature, noquench
magnetite is the least brittle iron oxide phase, which is pos-
itive with respect to scale integrity. However, an extensive
series of laboratory tests [4] has shown that the best scale
integrity occurs with oxide layers that largely exist of de-
composed wustite. For such scales, the test results are sub-
stantially better than for samples with predominantly com-
40 42 44 46 48 50
pact magnetite. Presumably, the toughness of the magnetite dll'rnlctlon angle (0 28)
matrix is enhanced by the presence of numerous tiny a-Fe 40 42 44 46 48 50
particles, and so is its resistance to cohesive failure. (b)
Moreover, the a-Fe particles precipitating near the scale- FT=760'C
steel interface connect to the steel base. Looking at sub-mi- CT=620'C
coilcentre
cron scale, this leads to a very rough interface, and conse- with quench
quently also to an increased resistance to adhesive failure of
the oxide scale.

HTXRD experiments
High Temperature X-Ray Diffraction (HTXRD) was used
40 42 44 46 48 50
to monitor the composition of scale in situ during simulat- dll'rnlctlon angle (" 28)
ed hot rolling and coil cooling conditions. Each HTXRD- 40 42 44 46 48 50
experiment comprised the following steps: (e)
1. A cold rolled Ti,Nb-IF steel strip (1ooxlOxl mm) was FT=760'C .
CT=380'C .
placed in Philips X'pert MPD diffractometer, equipped coilcentre
with a filtered Co-K, radiation source, a Raytech 12° po- noquench
sition sensitive detector (PSD) and an Anton Paar
HTK2000 heating stage.
i
2. The sample was heated in a Nz flow to a chosen simulat-
ed hot rolling temperature.
3. The sample was oxidised in a humid air flow (dewpoint
f
+20°C), while subjected to a time-temperature-cycle sim-
40 42 44 46 48 50
ulating finishing hot rolling and ROT cooling, to a simu- dll'rnlctlon angle (0 28)
lated CT.
4. The sample was slowly cooled down in a simulated coil
Figure 6. Examples of series of HTXRD patterns recorded during
cooling time-temperature cycle. A N z flow was supplied experiments simulating ferritic finishing rolling at 760°C, coiling at
when the centre of a coil was simulated, and a 620 or 380°C, and subsequent coil cooling (centre position).
99%Oz/1 %N z flow when cooling of a near-edge region a) CT=620°C, no quench. b) CT=620°C, quench simulated, 1h after
was simulated. coiling. c) CT=380°C, no quench.

402 steel research into 75 (2004) No.6

 Metal Forming - Hot Rolling

From the HTXRD experiments the following conclusions
could be derived:
- When tertiary scale starts to grow (after secondary
descaling), first only wustite is present. This is very ben-
eficial with respect to the deformability in the roll bite,
and probably the main reason why a good surface quality
on hot rolled strip is possible anyway. It takes some time
before magnetite starts to develop, but once this has hap-
pened, the fraction of magnetite increases rapidly, so that
the scale consists of substantial fractions of both wustite
and magnetite at the end of the ROT. Normally there is no
time to develop hematite up to this stage, except when the
rolling speed is rather low (thick gauges) in combination
with relatively high FTs and CTs, or blisters or cracks oc-
cur in the scale (e.g. during a rolling pass), resulting in
parts of the scale being cut off from the flux of outward-
ly diffusing ferrous ions. Wustite in these scale parts is
rapidly oxidised to magnetite and eventually hematite.
- Immediately after coiling at a high temperature
(CT>600°C; e.g. Fig. 6a and 6b), Fe diffuses from the
steel base into the scale. This leads to conversion of mag-
netite into wustite, and transformation of a-rich wustite
into a more Fe-rich type of wustite. The rates of these
(diffusion-controlled) processes increase with (coiling)
temperature. Near the edges, these processes are counter-
acted by the influence of air ingress between the coil
windings in combination with a faster temperature decay.
- Wustite decomposition into magnetite and ferrite, accord-
ing to reaction (9), proceeds rather slowly just below
570°C, accelerates with decreasing temperature, and be-
comes very fast around 400°C. When temperature is fur-
ther lowered, reaction (9) slows down, but does not stop
until the temperature is below 300°C.
Trendlines for scale composition
Figure 7 shows trend-lines of the scale composition as a
function of the coiling temperature, for the coil centre (top
diagram) and the coil edges (bottom diagram). These trend-
lines could be drafted on the basis of the HTXRD trials and
a large set of microscopy analysis results of scale on hot
rolled steels [4]. The strip head or tails can be regarded as a
special case of an edge.
Figure 7 may only be used for rough estimations of scale
compositions, as secondary influence factors are not taken
into account (such as alloy composition, mill layout, rolling
speeds, FT, interstand and ROT cooling practice).
Scale on commercial hot rolled strips
Empirically observed scale thicknesses and micro-struc-
tures can be better understood with the insights outlined
above. Figure 8 shows some examples. Figure 8a shows
the typical double-layered microstructure of scale on hot
rolled Dual Phase (DP) steel. Due. to the very low CT of
these steels (CT<200°C), no changes occur after coiling;
the scale structure is the same as at the end ofthe ROT. The
scale has an inner layer of retained wustite, which gives the
scale a low cohesive strength, and an outer layer of compact
magnetite. The pickling rate of such a double-layered scale
can be strongly enhanced by creation of cracks in the scale
by a tension leveller, since the inner layer is much easier to
100
90
e 80
i
.....
0: 70

- No wustite can be retained when a strip is coiled in the

most critical temperature region around 400°C. Wustite
.
E 80
II
50

retention is possible in two cases: either with a very low ]

.
'l5
40

30
coiling temperature (e.g. with hot rolled Dual Phase 0: 20
e
steels, which are coiled below 200°C), or with relatively W 10
01:
high coiling temperatures (compare Figures 6a and 6c). 0
100 200 300 400 500 600 700 600
- After coiling at a high coiling temperature (CT>700°C), Coiling temperature rq
more wustite is retained than with somewhat lower coil-
ing temperatures (e.g. 600-700°C), because the decompo- I coll.dges I l
sition kinetics of Fe-rich wustite are considerably slower
than those of a-rich wustite. The occurrence of wustite
crystals with varying Fe/Ocratios was evidenced by
wustite diffraction peak splitting and/or shifting during
HTXRD experiments.
- Wustite retention is strongly enhanced by coil quenching
after coiling at high temperatures. In contrast, quenching
has little or no effect on scale composition if CT<600°C.
- Substantial amounts of hematite are formed near the
__ hemalite _ compact magnelde
edges of coils coiled at high temperatures (CT> 500°C).
-&-decomposed wOstile - no quench - : - retained wOstile - no quench
In summary, coiling temperature and position within the __ decomposed wOstile - quenched coil __ retained wOstite·quenched coil

coil have been identified as the primary influence factors

with respect to the final composition of tertiary scale. For Figure 7. Trend-lines giving a rough indication of scale composition
high CT's, coil quenching significantly affects scale com- as a function of coiling temperature. Top: coil centre. Bottom: coil
position, compared to natural coil cooling in air. edges.

steel research into 75 (2004) No.6 403

Metal Forming - Hot Rolling
pickle than the outer layer. The scale structure at the edges
is the same as at the strip centre. However, the inner wustite
layer is thinner at the edges than it is at the centre. This also
applies to the total scale thickness, as expected from the low
CT (4.5 urn versus 5.5 urn for this particular coil).
The scale shown in figure 8b largely consists of decom-
posed wustite, and some areas of compact magnetite, due to
the coiling temperature of 450°C, which favours complete
wustite decomposition. This appeared to be beneficial for a
good scale integrity. The scale is slightly thicker (6 urn)
than in figure 8a (5.5 urn) due to the somewhat higher Fr.
Figures 8c and 8d present the microstructures of centre
and edge scale, respectively, on a coil that was quenched in
a water basin, after coiling at a rather high temperature
(CT=75I "C). A large fraction of retained wustite is present
at the centre of this coil (figure 8c), due to the combination
of quenching and a high CT. In contrast, the near-edge re-
gion shows an outer rim of hematite, a thick layer of com-
pact magnetite, and some decomposed wustite near the in-
terface (figure 8d). Obviously, the edges severely suffered
further oxidation after coiling. Note that the edge scale is
relatively thick (-10 urn) due to the high CT, whereas the
centre scale is much thinner (-6 urn) due to the low Fr.
From table I it can be understood why this hot rolled strip
exhibits a rather bad scale integrity over the whole strip
width. At the centre, this is due to the high fraction of re-
tained wustite, and at the edges due to the subtantial
amounts of hematite and compact magnetite. This is aggra-
vated by the thickness. However, the edge scale is much
more difficult to pickle than the centre scale, due to these
differences in scale structure and thickness.
Figure 8. Examples of optical microscopy images of scale on vari-
ous hot rolled steels (etching: 0.5% HCI). a) Dual Phase steel,
FT=850°C, CT=150°C, strip centre. b) Low carbon steel, FT=875°C,
CT=450°C, strip centre. c) Low carbon steel, FT=773°C,
CT=751DC, coil quenched, strip centre. d) same coil, strip edge.
404
Concluding remarks
A consistently good surface quality and downstream
processability of hot rolled strips requires control of tertiary
scale properties throughout the whole process route be-
tween secondary descaling and coil cooling to room tem-
perature.
In this respect, scale composition is more important than
scale thickness.
Finishing temperature (Fl') is the main parameter deter-
mining the final scale thickness at the strip centre, followed
by gauge and coiling temperature (CT). At the edges, CT is
more important.
In the roll bite, wustite is the preferred component. Fur-
thermore, too thick or over-chilled oxide layers should be
avoided to prevent scale cracking. Therefore strip cooling
should be carefully balanced.
At room temperature, decomposed wustite is the pre-
ferred component for a good scale integrity.
Scale integrity at the coil edges is generally significantly
inferior to that at the coil centre, due to lower fractions of
decomposed wustite at the edges and, in case of a high CT,
significant Fez03 formation.
Very low or very high CTs are detrimental to scale in-
tegrity, due to high fractions of retained wustite. This is fur-
ther aggravated by quenching, after coiling at a high CT.
For the coil centres, there seems to be a broad optimum
CT range around 550°C. For the coil edges, the optimum is
much narrower, at an inferior level, and shifted to a lower
temperature.
Acknowledgement
Parts of the presented research were financially support-
ed by the ECSC.
(A2004024; presented in Dusseldorf on 17 June 2003)
Contact: Dr. P. Henk Bolt
Corus Research, Development & Technology
P.O. Box 10000
1970 CA JJmuiden, The Netherlands
References
[I] U.R. Evans: Reviews of Pure and Applied Chemistry. 5 (1955), No.1,
1/21.
[2] U.R. Evans: "Velocity of film growth" in: The Corrosion and Oxida-
tion of Metals: Scientific Principles and Practical Applications, E.
Arnold Publishers Ltd., London, 1960, Chapter XX, p. 819/59.
[3] H.F. Marston, R. Ratcliffe, P.H. Bolt, M. van Lankveld, V. Lanteri, G.
Leprince, M. Jarl, J. Niska, R. Klima, M. ROder: "Oxidation and De-
carburisation of High Carbon, Special and General Steels", ECSC re-
search project 721O.PR/019, Final Report EUR 20497, 2002.
[4] H. Bolt, G. Bourdon, X. Cornet, S. Ehlers, F. Friedel and F. Steinert:
"Investigation of the Formation, Constitution and Properties of Scale
Formed during the Finishing Rolling, Cooling and Coiling of Thin Hot
Strips", ECSC Research Project 721O.PR/153, Draft Final Report,
2003.
steel research int. 75 (2004) No.6